Multisensing Green Plant Body Monitoring by Electrical Impedance Analysis

In this thesis, internal/external electrical signalling, as well as the impedance of several in-vivo (plants) systems, is studied. A brief introduction of intrinsic (natural) intra-plant communications is the foundation of an elemental intra-plant communication framework (that consists of easy-to-obtain components) that achieves prominent results employing an S-OOK modulation scheme. The latter enables the idea of inter alia, collect and transmit plant's health information through it. For instance, a Heat Pulse Velocity Sensor sensing could exploit this principle by measuring Sap Flow and using the plant as a communication channel. Subsequently, the development of an automated impedance measurement scheme with a simultaneous monitoring system is discussed extensively for practical usability and future reference. The purpose of such a system is to serve as a laboratory suite that gathers in vivo (plants) systems impedance and related environmental variables data for further analysis. The full system is designed to be flexible so that the user could modify the measurement parameters as well as the employed monitoring devices. This system served to collect several data sets through multiple research on tomato and tobacco plants that helped to validate the laboratory suite. On each experiment, the goal was to emulate both watering and drought conditions on the involved specimens. Every measure had a duration that ranges from a week to even a month. In those measurements, some of the observed results are remarkable: for instance, watering events seem to affect impedance with a 'rebound' effect whereas ambient light adds a periodic pattern to impedance. Moreover, a drought indication seems to be a decreasing frequency shift in impedance angle, particularly of the minimum value, along with a continuous and sustained increase of impedance modulus. Also, the results suggest that some monitored variables, such as soil moisture, statistically cause variations in both modulus and angle impedance at certain frequencies. Finally, once establishing the analysis framework, some straightforward Machine Learning applications using the collected data to inquire in the usability of this approach for consequent research are performed. Within the group of monitored environmental variables, soil moisture predicts the best both impedance modulus and angle, at low frequencies, under relative RMSE and R2 statistical criteria. The latter illustrates the potentiality of this topic for future in-depth related works and the employment of impedance-based models for weather and plant health prediction purposes.